The world’s first artificially intelligent lawyer has arrived. Called Ross, and built on IBM’s famous cognitive computer called Watson, it has been “employed” by US firm Baker & Hostetler to work in its bankruptcy practice.
According to Futurism.com, Ross can “read and understand language, postulate hypotheses when asked questions, research, and then generate responses (along with references and citations) to back up its conclusions. Ross also learns from experience, gaining speed and knowledge the more you interact with it”. (http://futurism.com/artificially-intelligent-lawyer-ross-hired-first-official-law-firm/)
It’s not just lawyers who should be looking over their shoulders. All sorts of knowledge workers could see their employment prospects and livelihoods threatened by artificial intelligence, including journalists, accountants, portfolio managers, even surgeons and physicians.
With the aid of the internet and easy access to case law and its interpretation, fewer lawyers may be required to resolve a bankruptcy procedure. Fewer analysts may be required to value a company with the aid of Bloomberg data and its accompanying suite of programmes. Lasers directed by unerringly accurate robots may well help reduce the time in the operating theatre and the dangers of doing so.
What is the impact of these newly adapted technologies on productivity in the industry and the economy as a whole? Let’s use the example of the artificially intelligent bankruptcy lawyer above. The productivity of the lawyers in bankruptcy practice can be defined as the number of cases concluded divided by the number of lawyer hours billed to do so*. Presumably the number of lawyers (and legal hours billed) will decline with the aid of Ross. Thus the surviving lawyers working on bankruptcy law in a legal practice will have become more productive in the sense of an increase in the ratio (cases concluded/hours billed). Measuring productivity in this case seems a simple task.
It becomes much more difficult to measure the productivity of a service provider when real output is much trickier, and sometimes impossible, to measure. One would not wish to measure the productivity of an analyst, journalist, artist or inventor of a new video game by the number of words written and published or number of pictures painted or pixels injected. The quality of the work produced is surely more important than the quantity of output and “quality” is recognised in revenues generated. As is admitted by the calculators of productivity, it is impossible to measure the productivity of government officials, because it is not possible to measure how much they produce. All that can be measured is their employment benefits – an input. In the case of an author, composer, copywriter or game developer, only the value of the royalties they have earned can be measured – their revenue line, not the time spent writing the masterpiece. Measuring productivity requires that inputs and outputs can be independently measured, which is not always the case, especially for service providers.
However, looking at the example of the number of bankruptcy cases (which we would regard as an independent measure of output), what if the quality of advice has improved even as the numbers of hours billed declines? The advice may be superior with the aid of Ross’s deep memory bank. How would we adjust for this quality dimension in our measure of legal productivity? The question is apposite for the service sector generally, where computers and data management (and improved knowledge) have presumably enhanced the quality of service provided by lawyers, analysts and other knowledge professions, including the improved offering of physicians supported by bigger data and better statistics. If the quality of advice has objectively improved, then any hour of consulting service will be delivering more in real terms than a case handled in the same time say 10 years before. The output of the consultant will in effect have increased, even if the input of time is the same. But by how much is the leading question. The physicians may be seeing the same number of patients a day, charging them higher fees, but they (their patients) are likely to be living longer and better lives.
In cases like this we will not be comparing like with like, apples with apples or aspirins with aspirins, making any measure of real output and so productivity over time a very difficult exercise and one subject to significant errors in what is measured. For example, your medical insurance may well have become more expensive – or your cover reduced – but are you not getting a better quality of medical service in return? And exactly how much better? In the case of medical insurance, only what you are paying – not your additional benefits – will find their way into the official price indices.
A further aspect is the impact of improved quality on the broader economy. If the bankruptcy cases are resolved with less billable time spent in court and hence with a larger percentage of debt being recovered with reduced legal expenses, this would be a clear gain to the creditors. Creditors would be better off in real terms, with less spent on legal fees and earlier resolution of their claims, meaning that the creditors could spend more on other goods or services or save more. And lawyers competing with each other for work that has become less costly for them to supply, may well charge you less for their time. It is competition for extra revenue that turns lower costs into lower prices – even in the legal profession – provided they do not collude on fees.
Could the GDP deflator, the price index that converts estimates of GDP in money of the day into a real equivalent, hope to pick this up with a high degree of accuracy? Enough to provide accurate measures of GDP or productivity growth over extended periods of time? The deflator used to convert nominal GDP into real GDP, attempts to adjust for quality improvements in the output of goods and, especially, services produced. Yet in South Africa, 68% of all value added is comprised of services of one kind or another the quality of which may well be changing over time, in ways that are very difficult to measure.
Ours is more of a service economy, than one that produces goods, the output of which is much more easily measured in units of more or less constant quality – for example number of bricks or tons of cement or steel. Thus, if we are underestimating quality improvements in the large service sector, we will be overestimating inflation and so underestimating the growth in real incomes, output and productivity.
Your real incomes and your productivity may well have increased even if you are taking home no more pay or other employment benefits. You may be benefitting from an enhanced quality of service as well as a very different mix of services than was available 10 years before, for example easy internet access that has so changed the way we work and play. This has become a particular problem in the developed world where prices as measured are generally falling. Deflation, rather than inflation, is the greater concern and nominal wages are not rising, even if productivity and the standard of living, differently measured and quality enhanced, is improving (though perhaps poorly recognised, as voters in their frustration at their constant money incomes turn to populists who promise a better standard of living). A mere one or two per cent extra a year factored into GDP or productivity growth measures, well within a range of possible measurement errors, would provide a very different impression of how the developed world is doing. A rising real standard of living, if only we could measure it, might well be accompanying stagnant employment benefits, when calculated in money of the day.
*One of the criteria the World Bank uses for measuring the ease of doing business in any country is outcomes in the bankruptcy courts. The tables below measure ease of doing business across a number of categories, and we show the SA and Australia findings where SA compares quite poorly. The ranking is, for example, 120/189 for ease of starting a business compared to 11 for Australia; and 41 for bankruptcy proceedings compared with 14 for Australia. Both countries rank poorly for trade across borders (130 and 89). Note we do much better than Australia when it comes to protecting minority investors: ranked 14 vs 66; and worse for getting credit, 59 Vs 5.
